The Agentic Shift is Upon Us
In the tech world, “dogfooding” is the ultimate litmus test. It’s common to test a product with internal employee users before that product ever touches the public. I did this for years at Google and I’m doing it now at Snowflake (with some really cool products that I hope will see the light of day soon)!
Recently, reports emerged that Google is dogfooding a product codenamed “Remy”. It’s supposedly not just another chatbot; it’s a proactive, 24/7 digital operator powered by Gemini. I’ve been testing a (probably) light version of this through Google Labs’ “CC” agent, which sifts through my Gmail, Calendar, and Drive to hand me a prioritized morning briefing.
But Remy represents something much bigger: the transition from an AI chatbot experience that is reactive and that answers to an AI chatbot experience that is proactive, that takes action without prompting.
So… why is this such a big deal?
Google has 180,000 employees. If even 25% of those employees dogfooded any of their products, they’d get a large swath of testers to provide a significant internal feedback loop to be able to iterate on a product pretty quickly. Testing agentic experiences with thousands of users and gaining valuable feedback is a pretty good starting point to evaluate user behavior.
This is in line with what Sundar Pichai noted in Google’s Q1 2026 earnings call, where he said they are focusing on three things: Intelligence, Agents and Agentic Coding. The speakers on the call noted that they have an opportunity to “go deeper” for people searching, to bring agentic workflows for consumers in search.
What is an agentic workflow? IBM defines an agentic workflow as “AI-driven processes where autonomous AI agents make decisions, take actions and coordinate tasks with minimal human intervention.”
People are already using agents: Nearly 50% of people claimed to have used an agent to help them shop in 2025. If people are doing deep research, or calling a call center, or chatting with a chatbot in their banking app – they are interacting with an agent. We know when we are explicitly using an agent, and in other circumstances, we probably don’t even realize when we are using an agent. They tend to be everywhere these days.
So, agents aren’t new to us. But, it does feel like if Google is dogfooding agentic experiences across Google products, once they launch Remy to the world, we’ll likely see consumer behavior change as they realize they can move from a reactive chatbot experience to assigning tasks and waiting for Google to execute upon them – proactively.
It’s not impossible to imagine a world where Google might take in context to know, for example, my preferences and routines and then take a proactive action.
Based on my own habits, a proactive agent would see my Calendar and email confirmations and note that I play padel once a month on Saturday afternoons. It wouldn’t wait for me to search “padel hub” or “padel courts near me” (like I did recently in Egypt). Instead, it would look at my open Saturdays for the next two months and suggest (or even just outright book) the court for me.
It could notice that I play tennis every Sunday at 10am from May to October and proactively renew my 8-week registration before the spot fills up.
It could even sync with my Blue Bottle receipts in my Gmail to recognize that I go there on the same day each week, and then proactively order my coffee and ensure it is waiting for me the moment I walk in every Tuesday morning.
These might seem like small conveniences, but they represent a fundamental pivot. Because of its sheer number of users, Google has the pole position to move us from a “search-and-click” society to a “delegate-and-done” society.
I continue to write this: Google has the opportunity to change behavior. With 260 million active users in the United States, if even 25% of them were to start using Google’s “Remy” (once – or if – it goes live), that means 65 million people will change their behavior and habits and potentially make a transition from simply asking questions of Google to leveraging Google’s agentic solutions for a more proactive interaction with the world.
Interactions with websites, web forms, booking agents, etc. will all change. Remy might be able to book a reservation for a restaurant or a hotel. It might be able to fully book a vacation – flights, hotels, etc. It will likely be able to research and purchase products agentically.
And at some point, there might be (and the operative word here is might) the opportunity to research healthcare symptoms and diagnoses as well as book healthcare appointments. Consider what searching for a doctor might look like when someone has agentic expectations…
The more people become accustomed to the agentic way of life, the more they will come to expect these experiences in their healthcare interactions.
Right now, it’s estimated that 50% of website traffic is now coming from a machine. What happens when this starts to approach 75%? The more machines take over visits to websites, the more it will be imperative to construct experiences that machines (rather than humans) can interact with. We need to be prepared for this future.
We need to pay attention to any time Google makes large scale changes that can impact consumer behavior. These changes have a trickle-down effect on everything we do and plan to do. Consumer behavior doesn’t exist in a silo. Once someone becomes accustomed to an agent booking their flights and coffee, they will expect that same agent to research their symptoms, find a specialist, and book an appointment in their EHR.
The question for every marketing leader is no longer “How do I show up in a search result?” It is: “Is my website readable by a machine?” And, “can an agent take action?”
Yes, the AI Overview is important. But don’t forget to think about what the future could look like when a machine (not a human) is accessing your website. These considerations mean doing more than a simple optimization. It means understanding things like:
Data Readiness. Where does my data come from and how do I present it to a machine? Is it structured to be consumed by APIs? Does my EHR (e.g., Epic) allow me to access data via an API to tap into the data that might allow a machine to take an action (e.g., book an appointment)?
Action Lists. Have I identified the top 10 actions (booking, refills, records, etc.) a patient would want an agent to do?
Interoperability. Does my EHR (like Epic) actually allow a machine to “talk” to it to close the loop on a task?
Industry Influences. Have I thought about how people are interacting agentically in other industries and how this could translate to healthcare?
Technology changes, and consumer behavior follows. Whether it’s retail, hospitality, or healthcare, the “agentic way of life” is coming. It’s time to stop optimizing for clicks and start building for the machines that will soon be doing the clicking for us.
It’s never a dull moment, is it?